Design Of Automatic Tracking And Video Recording System For Airports Drug Detection Dogs

Recently our country already has a relatively mature technical system for drug detection dogs,but there are still some problems such as high labor input cost and low training efficiency.Through the research and analysis of the current situation of the intelligent system of drug detection dog training at home and abroad,combined with the current actual situation in my country,this thesis proposes an automatic tracking and video recording scheme for airport drug detection dogs based on the fusion positioning of UWB and machine vision.The scheme takes the smart car as the platform,uses the image acquisition unit to obtain the real-time monitoring picture and video information of the drug detection dog training site and preprocesses it,and uses the deep learning algorithm to complete tasks such as target detection,tracking and positioning.The corresponding relationship is established to realize the accurate calculation of the dog body position.At the same time,combined with the UWB positioning system,the movement data of drug detection dogs can be quickly obtained without affecting the normal work of police officers,which further provides a basis for later data analysis.The experimental results show that the method proposed in this thesis can better realize target recognition and tracking,and has high practical value.The target recognition accuracy rate is 93.51%,the recall rate is84.91%,and the false alarm rate is only 0.32%.The positioning accuracy of the system is0.50 cm,and it has good robustness even when the environment is complex,which can meet the requirements for positioning accuracy in practical applications.The main research contents of this thesis include:（1）An intelligent judgment algorithm for the recognition of drug detection dogs training process is proposed.Taking advantage of the dog’s low sensitivity to red,the red suitcase is selected as the container for hiding drugs in daily training.Therefore,the system needs to identify the red suitcase,and then automatically determine the training results of the drug detection dog.This thesis firstly conducts a comparative study on different target detection algorithms;establishes an image data set about red suitcases,improves the target detection algorithm for the defect that the YOLOv3 tiny algorithm has insufficient recognition accuracy for red suitcases,increases the residual network model structure,and replaces the activation.The function is the Swish activation function,and the loss function is replaced by GIOU;the improved YOLOv3 tiny algorithm is trained and evaluated;the evaluation results show that the improved YOLOv3 tiny algorithm not only ensures light weight,but also improves the accuracy rate,and realizes the detection of red luggage.identification of the box.Then write an automatic judgment algorithm to output the overlapping area between the red suitcase and the drug detection dog and the duration of the status,so that the system can determine whether the drug detection dog has identified drugs.（2）Proposes a fusion localization tracking algorithm based on UWB and machine vision.This paper adopts KCF-based vision tracking algorithm,using the depth camera to acquire the dynamic coordinates of the drug-sniffing dog in the filming scene;based on the actual situation of the execution accuracy of the vision servo system and susceptibility to environmental noise and external interference,etc.,the design scheme of UWB-assisted localization is proposed.The system acquires coordinate data from both the UWB module and the depth camera.The UWB module provides the system with the dynamic coordinates of the drug-sniffing dog in a global application scenario;while the depth camera provides the system with the coordinates of the drug-sniffing dog’s movement in a local environment.（3）Build an automatic tracking video system for drug detection dogs,and conduct system tests around drug detection dog training videos.Firstly,the communication settings and model deployment of each module between the systems are completed,and the frame of the automatic tracking and video recording system for airport drug detection dogs is built;secondly,according to the structural requirements of the system,the modules of moving target recognition,tracking,data acquisition and processing are analyzed in detail.Finally,the research and development of system software is completed based on ROS and server.The function test and performance evaluation of automatic tracking video were carried out in the simulation scene of drug detection dog training,which verified the effectiveness and practicability of the constructed prototype system.